Metal-organic frameworks (MOFs) are crystalline porous materials. They have attracted much attention because of their characteristics such as well-defined porosities, high surface areas, and chemical stabilities [1, 2]. MOFs are promising materials for various technological applications, including gas storage [3], separation [4], drug delivery [5], and chemical sensing [6]. In recent years, MOFs have been shown to have potential applications as heterogeneous catalysts [7-9].

The efficient separation and recycling of catalysts is a major objective of sustainable and green chemistry. Heterogeneous catalysts have a number of significant advantages over homogeneous ones because they are easily recoverable, reusable, and waste is minimized. Filtration or centrifugation are widely used to separate solid catalysts from the liquid phase, but these methods are cumbersome, especially when the catalyst particle sizes are in the sub-micrometer to micrometer range; these are more active than bulk catalysts because mass transfer is less limited [10]. Magnetic separation of catalysts is an attractive alternative to filtration or centrifugation.

Recently, many attempts have been made to integrate MOFs with magnets to facilitate their separation using an external magnetic field [11-16]. Ke et al. [12, 14] reported the preparation of a novel type of porous core-shell microspheres with a magnetic core and a designable MOF shell using a versatile step-by-step assembly strategy; the porous Fe₃O₄@MIL-100 had good catalytic activity and recyclability in Claisen-Schmidt condensations. Although considerable progress has been made in the synthesis of such magnetic nanohybrids, their use as heterogeneous catalysts still needs to be explored. Recently, Zhang et al. [15, 16] reported the preparations of Fe₃O₄@IRMOF-3 and Fe₃O₄@Cu₃(BTC)₂-NH₂ core-shell nanohybrids and their catalytic performances in Knoevenagel condensations. These nanohybrids were active and selective, but their structures were damaged during recovery after the reaction. It is therefore necessary to select stable MOFs as the shell. Numerous MOFs have been reported to date; of these, UiO-66 and its amino-substituted analogue UiO-66-NH₂, synthesized using ZrCl₄ as the metal source, are highly stable [17-20]. The synergistic action of the Lewis sites and base groups in UiO-66-NH₂ is needed to achieve optimum performance in Knoevenagel condensation and cross-aldol reactions [20, 21].

Recently, Fe₃O₄@UiO-66-NH₂ was prepared using a step-by-step method by dispersing Fe₃O₄ consecutively in dimethylformamide (DMF) solutions of ZrCl₄ or 2-aminobenzenetricarboxylic [22]. However, the powder X-ray diffraction (PXRD) patterns of the resulting Fe₃O₄@UiO-66-NH₂ showed only a broad peak between 5° and 10° because of the poor degree of crystallinity of UiO-66-NH₂. Activity loss was observed during recycling experiments because of partial destruction of its porosity.

In this study, Fe₃O₄@UiO-66-NH₂ core-shell nanohybrids were prepared using a simpler and more efficient method and were evaluated as heterogeneous catalysts for Knoevenagel condensations. A schematic diagram of the procedure is shown in Scheme 1. The core-shell nanohybrids were highly active and selective in Knoevenagel condensations because UiO-66-NH₂ is bifunctional and mass transfer in the shell is less limited. The core-shell hybrid material Fe₃O₄@UiO-66-NH₂-3 prepared using three assembly cycles with well-defined structure and high porosity showed good recycling stability. It was recovered magnetically and reused at least four times without significant loss of catalytic activity and framework integrity.

As shown in Scheme 1, the magnetic nanohybrid was prepared using a step-by-step assembly method. The first step involved modification of Fe₃O₄ nanospheres with polyacrylate sodium salt (PASS), and the second step was assembly of Fe₃O₄@PASS nanospheres and MOF precursors. The Fe₃O₄ nanospheres were synthesized using a previously reported solvothermal method [16]. In a typical synthesis, the prepared Fe₃O₄ (1 g) was added to an aqueous solution of PASS (M_r=3000, 1.6%, 60 mL) and the mixture was shaken for 24 h. The solid product (denoted by Fe₃O₄@PASS) was collected using a magnet and washed with distilled water and DMF. Fe₃O₄@PASS (0.35 g) was dispersed in a DMF solution (80 mL) containing ZrCl₄ (20 mmol) and 2-aminobenzenetricarboxylic acid (20 mmol). The solid was collected magnetically after stirring for 4 h at 120 ℃ and then redispersed in a fresh solution containing the two reagents in the concentrations indicated above. After one to three assembly cycles, the samples were collected magnetically, washed with ethanol, and dried under vacuum at 100 ℃. The resultant nanohybrid materials were denoted by Fe₃O₄@UiO-66-NH₂-n, where n is the number of assembly cycles.

PXRD patterns were recorded using a Bruker D4 Endeavour with Cu K_α radiation (λ=0.15405 nm) at 40 kV and 40 mA. Infrared (IR) spectra were recorded using a Thermo Nicolet Nexus 470 Fourier-transform IR spectrometer. Scanning electronic microscopy (SEM) was performed using a Quanta 3D FEG instrument. Transmission electron microscopy (TEM) was performed using a Tecnai G2 F20 instrument at an accelerating voltage of 200 kV with a LaB6 filament; samples were prepared by placing a drop of a colloidal solution on a carbon-coated copper grid and evaporating the solvent in air at room temperature. Nitrogen sorption isotherms were recorded at -196 ℃ using an ASAP 2000 system in static mode. The samples were out-gassed at 120 ℃ for 3 h prior to the measurements.

Knoevenagel reactions were performed in a 25-mL round-bottomed flask equipped with a magnetic stirrer. In a typical experiment, a desired amount of catalyst (0 or 3% -NH₂) and benzaldehyde (8 mmol) were mixed in DMF (5 mL). The flask was placed in an oil bath that had been preheated to 80 ℃. The reaction was started by adding ethyl cyanoacetate (7 mmol) to the flask. The reaction was monitored using a gas chromatograph (GC-2010 plus, Shimadzu) equipped with an Rxi@-1 capillary column (30 m × 0.25 mm × 0.5 μm) and a flame ionization detector. The temperature program for gas chromatographic analysis was set as follows: the temperature was held at 150 ℃ for 5 min, then raised to 180 ℃ at 30 ℃/min and held for 15 min. The inlet and detector temperatures were 280 ℃. The analysis was performed directly after sampling to avoid any additional conversion. In the recycling experiments, the catalyst was recovered magnetically when the reaction was finished, washed with DMF (3 × 5 mL), and reused directly without further purification for the next run with fresh benzaldehyde and ethyl cyanoacetate. The catalyst was used for four consecutive runs.

Magnetic nanohybrids were prepared using a self-assembly method by dispersing magnets in a DMF solution containing two MOF precursors, i.e., ZrCl₄ and 2-aminobenzenetricarboxylic acid. Polyacrylate sodium chains in this composite acted as a bridge, connecting the internal Fe₃O₄ and external UiO-66-NH₂ layer. The crystallinities and compositions of the synthesized materials were investigated using PXRD; the results are shown in Fig. 1. The Fe₃O₄@PASS pattern had intense reflections at 2θ=30.1°, 35.5°, 43.1°, 53.4°, and 57.0°; these are typical of Fe₃O₄ (JCPDS 19-0629) and correspond to the (220), (311), (400), (422), and (511) planes, respectively [12, 22]. Fe₃O₄@UiO-66-NH₂ prepared using two or three cycles had additional intense reflections at 2θbetween 5°-60°, which are typical of UiO-66-NH₂ and confirm the formation of highly crystalline UiO-66-NH₂ on the magnetic nanoparticles [17, 18]. However, in our previous work, only a broad peak between 5° and 10° was observed in the PXRD pattern of Fe₃O₄@UiO-66-NH₂ prepared by dispersing magnets consecutively in DMF solutions each containing only one of the MOF precursors [22].

The PXRD results showed that the nanohybrids consisted of Fe₃O₄nanoparticles and UiO-66-NH₂, but we needed to determine whether the nanohybrids were Fe₃O₄@UiO-66-NH₂ core-shell nanohybrids or simply a physical mixture of the two separate phases, i.e., Fe₃O₄and UiO-66-NH₂. The microstructures of the synthesized Fe₃O₄@UiO-66-NH₂ nanohybrids were therefore examined using SEM and TEM; the images are shown in Fig. 2. Representative SEM images (Fig. 2(a)-(c)) show that the prepared Fe₃O₄@UiO-66-NH₂ nanohybrids were spherical and had narrow size distributions. The TEM images of the nanocomposites in Fig. 2(d)-(f) clearly show core-shell structures, with Fe₃O₄ as the core and UiO-66-NH₂ as the shell. The thickness of the porous MOF increased with increasing number of assembly cycles. The SEM-energy-dispersive X-ray spectroscopy (EDS) results are shown in Table 1. The weight percentages of UiO-66-NH₂ in the Fe₃O₄@UiO-66-NH₂ nanohybrids were calculated to be 14.1% (one cycle), 52.4% (two cycles), and 79.9% (three cycles), based on the atomic percentages of Zr and Fe. The amino group concentrations were 0.48 mmol/g (one cycle), 1.79 mmol/g (two cycles), and 2.73 mmol/g (three cycles). It can be concluded that well-defined crystalline UiO-66-NH₂ covered the Fe₃O₄ nanoparticle surfaces and the shell thickness could be controlled by the number of assembly cycles.

Nitrogen adsorption-desorption isotherms (Fig. 3) were used to verify the porous structures of the materials and to calculate their surface areas and pore volumes. All the Fe₃O₄@UiO-66-NH₂ samples gave adsorption-desorption isotherms intermediate between type I and type IV, indicating the copresence of micropores and mesopores in the materials [14, 22]. The Brunauer-Emmett-Teller (BET) surface areas and pore volumes are summarized in Table 2. The surface areas and pore volumes of Fe₃O₄@UiO-66-NH₂-n increased with increasing number of assembly cycles. The BET surface area increased from 44 to 342 m²/g, and the total pore volume increased from 0.11 to 0.31 mL/g. For comparison, a composite was also prepared using the step-by-step method described by Ke et al. [12, 22], in which the two components of UiO-66-NH₂ were added separately. The BET surface area of the composite was only 121 m²/g, and the total pore volume was 0.11 mL/g, even after 10 steps. The method in which the two components of UiO-66-NH₂ are added together is clearly more efficient for core-shell nanocomposite production.

The catalytic activities were tested by performing liquid-phase Knoevenagel condensations of benzaldehyde and ethyl cyanoacetate in the presence of Fe₃O₄@UiO-66-NH₂ or nano-sized UiO-66-NH₂. The results are summarized in Table 3. All the Fe₃O₄@UiO-66-NH₂ samples had high activities in this reaction. In blank experiments, yields less than 22% were obtained in 2 h (Table 3, entries 1 and 2). When nano-sized UiO-66-NH₂ (particle size less than 100 nm) was used, 80.3% of yield was achieved in 1 h, with a turnover frequency (TOF) of 26.8 h^-1. The activities of Fe₃O₄@UiO-66-NH₂ and nano-sized UiO-66-NH₂ (with equivalent amounts of UiO-66-NH₂) were compared under identical conditions. The TOFs and selectivities of all the core-shell magnetic MOFs were comparable to or higher than those of nano-sized UiO-66-NH₂.

Fig. 4 shows the time courses of product formation using various Fe₃O₄@UiO-66-NH₂-nnanohybrids. The shell thickness is clearly one of the crucial parameters that affect the specific activities. Fe₃O₄@UiO-66-NH₂-1, which has a thin shell ( < 50 nm), had the best activity (TOF=28.9 h^-1) under identical conditions. The activity decreased with increasing shell thickness and Fe₃O₄@UiO-66-NH₂-3, with the thickest shell (150-200 nm), had the lowest activity (TOF=24.4 h^-1); this is because the diffusion efficiency in the porous material was lower in a thick shell. The data in Table 3 show that although UiO-66 was much less active than UiO-66-NH₂, the yield achieved using the NH₂-free MOF was higher (42.3% after 1 h) than that achieved in the blank non-catalytic reaction. This suggests that the Knoevenagel reaction can be promoted by Zr (IV) centers (Lewis acid sites). Yang et al. [20] recently suggested that UiO-66-NH₂ was bi-functional. The high catalytic performance of the core-shell nanohybrid in the Knoevenagel condensation is probably a result of the bi-functional character of UiO-66-NH₂ and better mass transfer in the shells.

Once we had established that Fe₃O₄@UiO-66-NH₂ was a good catalyst for the Knoevenagel condensation between benzaldehyde and ethyl cyanoacetate, we extended the study to the condensations of various aromatic aldehydes with cyanoacetate. The results are listed in Table 4. The activity of Fe₃O₄@UiO-66-NH₂-3 in the reaction with 4-methylbenzaldehyde (Table 4, entry 2) was similar to that in the reaction with the unsubstituted benzaldehyde (Table 4, entry 1). The activities in the reactions with nitro-substituted benzaldehydes were much higher than that with benzaldehyde; ≥97% yields were achieved in only 10 min (Table 4, entries 3-5). This reflects the accelerating effect of the strongly electron-withdrawing nitro group in reactions involving nucleophilic attack at a carbonyl group. The catalyzed reactions of the three nitro-substituted benzaldehydes proceeded readily to give≥97% yields in 10 min, but the initial reaction rates of these substrates were different. The initial reaction rate with 4-nitrobenzaldehyde was lower than those with the 2-and 3-isomers; the 3-isomer gave the highest reaction rate. The effect of substitution on the reactivity of benzaldehyde is similar to that observed for the Knoevenagel condensation catalyzed by UiO-66-NH₂ [20].

The catalyst was also active in the Knoevenagel reactions of larger aromatic aldehydes (1-naphthaldehyde and 9-anthraldehyde) with cyanoacetate, although the yields were lower for bulky aldehyde substrates (Table 4, entries 1, 6, and 7). The decreased yields indicate size effects. Table 4, entries 1, 6, and 7 show that the yield from the reaction of 9-anthraldehyde with ethyl cyanoacetate (2 h, 32.4%) was much lower than that for 1-naphthaldehyde (2 h, 52.2%), which was much lower than that for benzaldehyde (2 h, 98.0%). The order of the activities is benzaldehyde (~6.0 × 4.3 Ų) > 1-naphthaldehyde (~6.0 × 7.2 Ų) > 9-anthraldehyde (~5.9 × 9.2 Ų). These results show that the probability of bulky substrates forming transition-state complexes is significantly decreased by the limited space in the porous catalyst.

In addition to catalytic activity and selectivity, long-term stability is important for solid catalysts. Although Fe₃O₄@UiO-66-NH₂-1 had the highest activity, Fe₃O₄@UiO-66-NH₂-3, with a higher amino loading, was used in recycling experiments for the Knoevenagel condensation between benzaldehyde and ethyl cyanoacetate in DMF. After each cycle, the solid was magnetically separated, washed with DMF, and then reused in the next cycle. Fig. 5 shows that the yields and selectivities in four consecutive runs were almost the same. However, in our previous work, activity loss was clearly observed when Fe₃O₄@UiO-66-NH₂ prepared by dispersing magnets consecutively in DMF solutions each containing one of the MOF precursors was used [22]. This new synthetic method clearly improves the recycling stability. We used PXRD and nitrogen adsorption-desorption isotherm measurements to verify the high stability. After four consecutive cycles, the peak positions and shapes of the PXRD patterns (Fig. 6) had not changed significantly, confirming that Fe₃O₄@UiO-66-NH₂-3 is a stable heterogeneous catalyst and the framework porous structures were unchanged during the reaction. The nitrogen adsorption-desorption isotherms (Fig. 7) show that the textural properties of the catalyst were unchanged by the reaction. SEM and TEM images (Fig. 8) showed no obvious differences between the sizes and morphologies of the catalyst before and after the reaction, again proving that the catalyst is stable during the reaction. A hot-filtration experiment was also performed to confirm the heterogeneous nature of the catalytic reaction (Fig. 9). The solid catalyst was removed, using a magnet, from the hot solution 0.5 h after initiating the catalytic test. The reaction of the filtrate was then monitored for another 2 h. No significant catalytic conversion was observed, indicating that no active sites leached from the solid.

Fe₃O₄@UiO-66-NH₂ with a well-defined structure was synthesized by dispersing magnets in a DMF solution containing two MOF precursors, i.e., ZrCl₄ and 2-aminobenzenetricarboxylic acid. Its core-shell structure was confirmed using TEM, its composition was determined using SEM-EDS, and the presence of amino groups was confirmed using IR spectroscopy. The sharp and intense diffraction peaks in the PXRD pattern of the shell of Fe₃O₄@UiO-66-NH₂ prepared using three assembly cycles of the new method showed that the crystallinity of its shell was higher than that of Fe₃O₄@UiO-66-NH₂ prepared by dispersing magnets consecutively in DMF solutions each containing one of the MOF precursors. The new material also had a higher surface area and pore volume. The new core-shell Fe₃O₄@UiO-66-NH₂ nanohybrid is an ideal recyclable catalyst for Knoevenagel condensations because of the bifunctionality of UiO-66-NH₂ and better mass transfer in the nano-sized shell. It is worth noting that the activities of the magnetic UiO-66-NH₂ materials were comparable to or higher than that of nano-sized UiO-66-NH₂. The recycling stability of the new material was improved by its high degree of crystallinity and high porosity. It was magnetically separated using an external magnet and was stable in terms of activity and structure when recycled at least four times in the Knoevenagel condensation of benzaldehyde and ethyl cyanoacetate in DMF. The effect of substitution on the reactivity of benzaldehyde and size effects of substrates were similar to those observed for the Knoevenagel condensation catalyzed by UiO-66-NH₂. These MOF-based core-shell magnetic catalysts, which can be prepared simply, are expected to have large-scale industrial applications, in which separation and recycling are important to reduce costs and waste production.